The present invention relates to techniques for searching a database, and more specifically, to name-searching in mixed-entity lists.
There are various techniques for searching for names in structured lists. Many structured lists contain two or more types of named entities, such as personal names, organization names, product names, place-names, and others. Searching such lists may require that the searching and matching logic be expressed in terms of properties common to all the named-entity types comprised within the list. For example, the common properties may include the characters and/or symbols in which they are represented, length, number of tokens and other such basic properties.
Existing search results typically focus on basic string-similarity and token-alignment techniques, which do not approach the intuitive outcomes produced by humans. In some cases these similarity-measurement techniques may be supplemented by use of standard stop-word and noise-word lists, to focus matching and scoring logic on the more meaningful constituent elements in each candidate match-name.